High density lipoproteins (HDL) are being recognized as a remarkably structurally and functionally heterogeneous group of particles in blood plasma. However, HDL is treated as a homogeneous lipoprotein type in epidemiological studies and in development and testing of treatments to reduce coronary heart disease (CHD). The concentration of HDL-cholesterol or its major protein apoA-I is a strong inverse risk factor for CHD. However, a consensus is developing that knowledge of the diversity of HDL in humans is needed to make progress on the meaning of a low or a high HDL level and how to use information on HDL to develop treatments that can lower CVD. HDL exists in a plethora of subpopulations defined by content of proteins that have diverse relations to risk of CHD. For example, the concentration of HDL that has apoC-III predicts increased rather than decreased risk of CVD, and HDL with apoC-III lacks protective functions possessed by the total HDL and HDL that does not have apoC-III. Concordance of abnormal function and increased CHD risk of certain HDL subtypes supports the key concept that apolipoproteins alter the function of HDL, and that this altered function is involved in effects of the HDL types on atherosclerosis and incidence of CVD. The first specific aim is to measure the plasma concentrations of HDL subtypes defined by their containing or not containing one or more of at least 60 proteins that have been found in HDL by proteomics. Each protein has a known function related to HDL or CHD such as effects on cells that participate in atherosclerosis; cholesterol transfer to and from cells and HDL; inhibiting oxidation or inflammation; hemostasis; and innate immunity. Each protein will define a pair of HDL subtypes, that either have or do not have a specific protein, e.g HDL with apoC-III, HDL without apoC-III. We will determine for each new type of HDL the range of its concentration and its stability over time in samples from representative populations. We will characterize the HDL types by lipid and protein contents and by size. We anticipate that data reduction will produce a panel of 20 or fewer pairs of HDL types for the second specific aim to evaluate for risk of CHD in two prospective US cohorts of women (Nurses' Health Study II) and men (Health Professionals Follow-up Study). In the third specific aim, we will select HDL types individually that have independent and strong relations to risk, and that add to the magnitude of relative risk, adverse or protective. We will devise a personalized risk index of HDL that combines the information on protective, nonprotective, and adverse types. We will also study indexes of HDL subtypes grouped according to the known major function of the protein, e.g. cholesterol transfer, hemostasis, oxidation, etc. This could be relevant to the particular etiology of a person's CHD. We will test how much the indexes built with novel HDL types together add to standard CHD risk prediction models. Our ultimate aim is to discover indexes that are useful in predicting CHD risk; evaluating response to diet and drug treatments; and discovering new specific HDL targets for development of drugs to prevent CHD.